The future is Africa: building AI-enabled EdTech that works for every learner
By 2050, one in three children on the planet will live in Africa. Yet more than 70% of children in low- and middle-income countries cannot read a basic text by age 10, and in Sub-Saharan Africa the rate was 86% even before the pandemic. That's a warning sign. Without faster gains in foundational skills, a young continent risks a shortfall in opportunity and jobs.
AI is entering classrooms everywhere. But most tools were built for high-income contexts with steady internet, rich datasets, and different classroom realities. If we don't build for local needs, the gap widens instead of closes.
AI works-when it's built with purpose
We're already seeing meaningful results. Rajasthan used AI-powered assessment to score paper worksheets for 4.5 million learners. In Kenya, nearly 400,000 children use EIDU, a structured pedagogy solution showing learning gains. In Edo, Nigeria, an after-school program saw significant progress in six weeks with AI tutoring plus teacher guidance. Big tech is moving too, from Gemini Guided Learning to Claude's learning mode and OpenAI's study mode.
But context matters. If a tool in rural Tanzania builds a lesson around pizza instead of chapati, it misses the mark. The goal is simple: practical tools that meet learners where they are and help teachers do their best work.
Three priorities to close the learning gap
1) Build equitably-AI that works everywhere
- Respect language and culture. Support local languages, relevant examples, and national curricula.
- Prioritize foundational reading and math. Keep instruction tight, explicit, and aligned to how children learn.
- Engineer for constraints. Low bandwidth, offline-first workflows, small models that run on basic devices.
- Fit real classrooms. Short, daily teacher prompts, printable formats, and easy data capture that doesn't add workload.
2) Work collaboratively-local developers, educators, governments, and big tech
- Co-design with teachers from day one. The best features come from real classrooms, not whiteboards.
- Share proofs, not just promises. Open up evaluation methods, safety checks, and content vetting.
- Grow local data responsibly. Only 0.2% of training data comes from Africa and South America-change that with ethical, curriculum-aligned datasets.
- Build smart partnerships. Examples are emerging across LMICs, including Rwanda, Kenya, and India, with clear roles and shared value.
3) Build evidence and quality-safe, effective, and scalable
- Set learning targets first. What reading and math gains should a tool deliver in 6-12 weeks?
- Test in real schools. Pilot, measure, compare to alternatives, and report results openly.
- Adopt benchmarks and guardrails. Independent checks on accuracy, bias, safety, and cost-to-impact.
- Share what works. Evidence efforts like EdTech Tulna help decision-makers separate signal from noise.
What good looks like in practice
Across LMICs, teams are proving what's possible: adaptive learning pilots in Côte d'Ivoire, The Gambia, and Mali; WhatsApp-based tutors in Ghana; teacher-focused solutions in Ethiopia; and youth skills programs in Tanzania and Mauritius. These projects are building the playbook for responsible adoption at scale.
For a broader view of the learning crisis, see the World Bank's overview of learning poverty: Learning Poverty.
A practical action plan for the next 6-12 months
For ministries and state agencies
- Publish a short AI-in-education roadmap focused on foundational learning, infrastructure, and teacher support.
- Approve low-bandwidth, offline-first standards for pilots and procurement.
- Stand up a national sandbox with data, curriculum mappings, and clear evaluation protocols.
- Fund 3-5 rapid pilots with independent measurement and public reporting.
For districts and school networks
- Choose one high-leverage use case: assessment, lesson prep, or tutoring. Start small, measure weekly.
- Train lead teachers as "implementation coaches" with simple checklists and office hours.
- Collect classroom-ready content in local languages and share it under safe licenses.
For developers and startups
- Build offline-first. Optimize for latency, memory, and battery life on low-cost Android devices.
- Ship with curriculum alignments, reading-level tagging, and clear teacher workflows.
- Audit outputs for bias and hallucinations; publish your test sets and failure modes.
- Report cost-per-learning-gain, not vanity metrics.
For funders and multilaterals
- Back shared infrastructure: datasets, benchmarks, and open reference implementations.
- Require independent evaluation and transparent reporting for grants and procurements.
- Support scale-up only after tools show repeatable gains across diverse contexts.
Momentum is building
In November 2025, leaders across education and technology met at the AI for Education Summit in Nairobi to focus on teacher support, personalized learning, and assessment. The message was consistent: solutions must be systemic, grounded in local realities, and aligned across actors-from classrooms to ministries.
The objective is clear: help every student move from foundational skills to job-relevant skills. Responsible AI can speed that path if we stay practical and evidence-led.
Where to go from here
If you build products, ship for low-resource classrooms first. If you lead systems, create the conditions for safe experimentation and fast learning. If you fund solutions, pay for evidence and the infrastructure everyone needs.
Educators who want structured upskilling in AI can explore role-based learning paths here: AI Courses by Job.
Africa's young learners can win-if the tools in their hands match the realities of their schools. Let's make that the standard, not the exception.
Your membership also unlocks:
